On 12/11/06, Steinar H. Gunderson <sgunderson@bigfoot.com> wrote:
> On Mon, Dec 11, 2006 at 11:09:13AM -0200, Daniel van Ham Colchete wrote:
> >> You know what? I don't.
> > So test it yourself.
>
> You're making the claims, you're supposed to be proving them...
>
> > As I said, it is an example. Take floatpoint divisions. You have
> > plenty of ways of doing it: 387, MMX, SSE, 3dNow, etc... Here GCC have
> > to make a choice.
>
> No, you don't. MMX, SSE and 3Dnow! will all give you the wrong result
> (reduced precision). SSE2, on the other hand, has double precision floats, so
> you might have a choice there -- except that PostgreSQL doesn't really do a
> lot of floating-point anyhow.
>
> > And this is only one case. Usually, compiler optimizations are really
> > complex and the processor's timings counts a lot.
>
> You keep asserting this, with no good backing.
>
> > If you still can't imagine any case, you can read Intel's assembler
> > reference. You'll see that there are a lot of ways of doing a lot of
> > things.
>
> I've been programming x86 assembler for ten years or so...
So, I'm a newbie to you. I learned x86 assembler last year.
> > Steinar, you should really test it. I won't read the PostgreSQL source
> > to point you were it could use SSE or SSE2 or whatever. And I won't
> > read glibc's code.
>
> Then you should stop making these sort of wild claims.
>
> > You don't need to belive in what I'm saying. You can read GCC docs,
> > Intel's assembler reference, AMD's docs about their processor and
> > about how diferent that arch is.
>
> I have.
>
> /* Steinar */
Okay, I'll do the benchmarks. Just sent an e-mail about this to the
list. If you have any sugestions of how to make the benchmark please
let-me know.
I like when I prove myself wrong. Although it's much better when I'm
right :-)...
Best regards,
Daniel Colchete